Convolutional Neural Networks¶
Project: Write an Algorithm for a Dog Identification App¶
In this notebook, some template code has already been provided for you, and you will need to implement additional functionality to successfully complete this project. You will not need to modify the included code beyond what is requested. Sections that begin with '(IMPLEMENTATION)' in the header indicate that the following block of code will require additional functionality which you must provide. Instructions will be provided for each section, and the specifics of the implementation are marked in the code block with a 'TODO' statement. Please be sure to read the instructions carefully!
Note: Once you have completed all of the code implementations, you need to finalize your work by exporting the iPython Notebook as an HTML document. Before exporting the notebook to html, all of the code cells need to have been run so that reviewers can see the final implementation and output. You can then export the notebook by using the menu above and navigating to \n", "File -> Download as -> HTML (.html). Include the finished document along with this notebook as your submission.
In addition to implementing code, there will be questions that you must answer which relate to the project and your implementation. Each section where you will answer a question is preceded by a 'Question X' header. Carefully read each question and provide thorough answers in the following text boxes that begin with 'Answer:'. Your project submission will be evaluated based on your answers to each of the questions and the implementation you provide.
Note: Code and Markdown cells can be executed using the Shift + Enter keyboard shortcut. Markdown cells can be edited by double-clicking the cell to enter edit mode.
The rubric contains optional "Stand Out Suggestions" for enhancing the project beyond the minimum requirements. If you decide to pursue the "Stand Out Suggestions", you should include the code in this IPython notebook.
Why We're Here¶
In this notebook, you will make the first steps towards developing an algorithm that could be used as part of a mobile or web app. At the end of this project, your code will accept any user-supplied image as input. If a dog is detected in the image, it will provide an estimate of the dog's breed. If a human is detected, it will provide an estimate of the dog breed that is most resembling. The image below displays potential sample output of your finished project (... but we expect that each student's algorithm will behave differently!).
In this real-world setting, you will need to piece together a series of models to perform different tasks; for instance, the algorithm that detects humans in an image will be different from the CNN that infers dog breed. There are many points of possible failure, and no perfect algorithm exists. Your imperfect solution will nonetheless create a fun user experience!
The Road Ahead¶
We break the notebook into separate steps. Feel free to use the links below to navigate the notebook.
- Step 0: Import Datasets
- Step 1: Detect Humans
- Step 2: Detect Dogs
- Step 3: Create a CNN to Classify Dog Breeds (from Scratch)
- Step 4: Use a CNN to Classify Dog Breeds (using Transfer Learning)
- Step 5: Create a CNN to Classify Dog Breeds (using Transfer Learning)
- Step 6: Write your Algorithm
- Step 7: Test Your Algorithm
Step 0: Import Datasets¶
Import Dog Dataset¶
In the code cell below, we import a dataset of dog images. We populate a few variables through the use of the load_files function from the scikit-learn library:
train_files,valid_files,test_files- numpy arrays containing file paths to imagestrain_targets,valid_targets,test_targets- numpy arrays containing onehot-encoded classification labelsdog_names- list of string-valued dog breed names for translating labels
from sklearn.datasets import load_files
from keras.utils import np_utils
import numpy as np
from glob import glob
def load_dataset(path):
"""
Load train, test, and validation datasets
Parameters:
path: a directory to store dataset
Return:
dog_files: an array that contains file paths of all dog files
dog_targets: an array that contains one-hot encode labels of all dog files.
"""
data = load_files(path)
dog_files = np.array(data['filenames'])
dog_targets = np_utils.to_categorical(np.array(data['target']), 133)
return dog_files, dog_targets
# load train, test, and validation datasets
train_files, train_targets = load_dataset('/data/dog_images/train')
valid_files, valid_targets = load_dataset('/data/dog_images/valid')
test_files, test_targets = load_dataset('/data/dog_images/test')
# load list of dog names
dog_names = [item[27:-1] for item in sorted(glob("/data/dog_images/train/*/"))]
# print statistics about the dataset
print('There are %d total dog categories.' % len(dog_names))
print('There are %s total dog images.\n' % len(np.hstack([train_files, valid_files, test_files])))
print('There are %d training dog images.' % len(train_files))
print('There are %d validation dog images.' % len(valid_files))
print('There are %d test dog images.'% len(test_files))
Visualize the First 20 Training Images¶
import cv2
import matplotlib.pyplot as plt
%matplotlib inline
def visualize_img(img_path, ax):
"""
Visualize the image that is stored in the file path
Parameters:
img_path: a string-valued file path to an image
ax: object or array of Axes objects
Return:
None
"""
img = cv2.imread(img_path)
ax.imshow(cv2.cvtColor(img, cv2.COLOR_BGR2RGB))
fig = plt.figure(figsize=(20, 10))
for i in range(10):
ax = fig.add_subplot(2, 5, i+1, xticks=[], yticks=[])
visualize_img(train_files[i], ax)
Import Human Dataset¶
In the code cell below, we import a dataset of human images, where the file paths are stored in the numpy array human_files.
import random
random.seed(8675309)
# load filenames in shuffled human dataset
human_files = np.array(glob("/data/lfw/*/*"))
random.shuffle(human_files)
# print statistics about the dataset
print('There are %d total human images.' % len(human_files))
Visualize the Training human Image¶
fig = plt.figure(figsize=(10, 5))
ax = fig.add_subplot(1, 1, 1, xticks=[], yticks=[])
visualize_img(human_files[0], ax)
Step 1: Detect Humans¶
We use OpenCV's implementation of Haar feature-based cascade classifiers to detect human faces in images. OpenCV provides many pre-trained face detectors, stored as XML files on github. We have downloaded one of these detectors and stored it in the haarcascades directory.
In the next code cell, we demonstrate how to use this detector to find human faces in a sample image.
def show_bounding_box_4_detected_face(img_path):
"""
Display bounding boxs for each detected face in an image
Parameters:
img_path: a string-valued file path to an image
Return:
None
"""
# extract pre-trained face detector
face_cascade = cv2.CascadeClassifier('haarcascades/haarcascade_frontalface_alt.xml')
# load color (BGR) image
img = cv2.imread(img_path)
# convert BGR image to grayscale
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
# find faces in image
faces = face_cascade.detectMultiScale(gray)
# print number of faces detected in the image
print('Number of faces detected:', len(faces))
# get bounding box for each detected face
for (x,y,w,h) in faces:
# add bounding box to color image
cv2.rectangle(img,(x,y),(x+w,y+h),(255,0,0),5)
# convert BGR image to RGB for plotting
cv_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
# display the image, along with bounding box
plt.imshow(cv_rgb);
show_bounding_box_4_detected_face(human_files[3])
def show_bounding_box_4_detected_face_pass_img(img):
"""
Display bounding boxs for each detected face in an image
Parameters:
img: a color (BGR) image has been read by OpenCV
Return:
None
"""
face_cascade = cv2.CascadeClassifier('haarcascades/haarcascade_frontalface_alt.xml')
# convert BGR image to grayscale
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
# find faces in image
faces = face_cascade.detectMultiScale(gray)
# get bounding box for each detected face
for (x,y,w,h) in faces:
# add bounding box to color image
cv2.rectangle(img,(x,y),(x+w,y+h),(255,0,0),5)
# convert BGR image to RGB for plotting
cv_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
# display the image, along with bounding box
plt.imshow(cv_rgb)
Before using any of the face detectors, it is standard procedure to convert the images to grayscale. The detectMultiScale function executes the classifier stored in face_cascade and takes the grayscale image as a parameter.
In the above code, faces is a numpy array of detected faces, where each row corresponds to a detected face. Each detected face is a 1D array with four entries that specifies the bounding box of the detected face. The first two entries in the array (extracted in the above code as x and y) specify the horizontal and vertical positions of the top left corner of the bounding box. The last two entries in the array (extracted here as w and h) specify the width and height of the box.
Write a Human Face Detector¶
We can use this procedure to write a function that returns True if a human face is detected in an image and False otherwise. This function, aptly named face_detector, takes a string-valued file path to an image as input and appears in the code block below.
def OpenCV_face_detector(img_path):
"""
Detect whether if any faces in an image
Parameters:
img_path: a string-valued file path to an image
Return:
True: if face is detected in image stored at img_path
False: if face is NOT detected in image stored at img_path
"""
face_cascade = cv2.CascadeClassifier('haarcascades/haarcascade_frontalface_alt.xml')
img = cv2.imread(img_path)
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
faces = face_cascade.detectMultiScale(gray) # face position (x, y, w, h)
return len(faces) > 0
def Improved_OpenCV_face_detector(img_path):
"""
Detect whether if any faces in an image
Parameters:
img_path: a string-valued file path to an image
Return:
faces: number of faces that have been detected in an image
img: a color (BGR) image has been read by OpenCV
"""
face_cascade = cv2.CascadeClassifier('haarcascades/haarcascade_frontalface_alt.xml')
img = cv2.imread(img_path) #BGR image
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
faces = face_cascade.detectMultiScale(gray)
return faces, img
(IMPLEMENTATION) Assess the Human Face Detector¶
Question 1: Use the code cell below to test the performance of the face_detector function.
- What percentage of the first 100 images in
human_fileshave a detected human face? - What percentage of the first 100 images in
dog_fileshave a detected human face?
Ideally, we would like 100% of human images with a detected face and 0% of dog images with a detected face. You will see that our algorithm falls short of this goal, but still gives acceptable performance. We extract the file paths for the first 100 images from each of the datasets and store them in the numpy arrays human_files_short and dog_files_short.
Answer:
100 percentage of the first 100 images in human_files with a detected human face. However, 11 percentage of the first 100 images in dog_files have a detected human face. It shows that the face detector did well on human images, but it missclassified some images on dog files as human face.
human_files_short = human_files[:100]
dog_files_short = train_files[:100]
def OpenCV_face_detector_performance(img_path):
"""
Test the performance of the face_detector algorithm on the images in human_files_short and dog_files_short.
Parameters:
img_path: a string-valued file path to an image
Return:
accuracy of face detection performance in percentage
"""
count = 0
for img in img_path:
if(OpenCV_face_detector(img)):
count += 1
return (count/len(img_path))*100
human_files_performance = OpenCV_face_detector_performance(human_files_short)
dog_files_performance = OpenCV_face_detector_performance(dog_files_short)
print("OpenCV face detector performace on human files is {:.0f}%".format(human_files_performance))
print("OpenCV face detector performace on dog files is {:.0f}%".format(dog_files_performance))
def show_misclassified_imgs(img_path):
"""
Display images that are missclassified by the face_detector algorithm in dog_files_short.
Parameters:
img_path: a string-valued file path to an image
Return:
None
"""
misclassified_imgs = []
for img in img_path:
faces, img = Improved_OpenCV_face_detector(img)
# create a image list of misclassified images
if(len(faces)):
misclassified_imgs.append(img)
print(len(misclassified_imgs), "dog images are misclassified as human face.")
# show misclassifed imgs
fig = plt.figure(figsize=(10, 5))
for i in range(len(misclassified_imgs)):
show_bounding_box_4_detected_face_pass_img(misclassified_imgs[i])
plt.figure(i+1);
plt.show();
dog_files_short = train_files[:100]
show_misclassified_imgs(dog_files_short)
Question 2: This algorithmic choice necessitates that we communicate to the user that we accept human images only when they provide a clear view of a face (otherwise, we risk having unneccessarily frustrated users!). In your opinion, is this a reasonable expectation to pose on the user? If not, can you think of a way to detect humans in images that does not necessitate an image with a clearly presented face?
Answer: It is not a reasonable expectation to pose on the user to provide a image with a clearly presented face. In some cases, people may need to shoot moving subjects or have no high resolution devices supported. Therefore, they might not be able to provide clear images all the time. We can preprocess our images by adding some noise to simulate the images which can not show clear faces such as low resolution, blurring, etc. Then train our network model to recognize those images with unclear faces. By this way, our face recoginition app will be more user-friendly.
We suggest the face detector from OpenCV as a potential way to detect human images in your algorithm, but you are free to explore other approaches, especially approaches that make use of deep learning :). Please use the code cell below to design and test your own face detection algorithm. If you decide to pursue this optional task, report performance on each of the datasets.
Another approach: Face Recognition in Python¶
## (Optional) TODO: Report the performance of another
## face detection algorithm on the LFW dataset
### Feel free to use as many code cells as needed.
import face_recognition
image = face_recognition.load_image_file(human_files[3])
face_locations = face_recognition.face_locations(image)
# load color (BGR) image
img = cv2.imread(human_files[3])
# convert BGR image to grayscale
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
# find faces in image
face_locations = face_recognition.face_locations(gray)
# print number of faces detected in the image
print('Number of faces detected:', len(face_locations))
# get bounding box for each detected face
for (top,right,bottom,left) in face_locations:
# add bounding box to color image
cv2.rectangle(img,(left,top),(right,bottom),(255,0,0),2) # (255,0,0) is bounding box color, 2 is bounding box line width
# convert BGR image to RGB for plotting
cv_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
# display the image, along with bounding box
plt.imshow(cv_rgb)
plt.show()
def face_recognition_detector(img_path):
"""
Detect whether of any faces in an image
Parameters:
img_path: a string-valued file path to an image
Return:
True: if face is detected in image stored at img_path
False: if face is NOT detected in image stored at img_path
"""
img = cv2.imread(img_path)
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
faces = face_recognition.face_locations(gray)
return len(faces) > 0
def face_recognition_detector_performance(img_path):
"""
Test the performance of the face_detector algorithm on the images in human_files_short and dog_files_short.
Parameters:
img_path: a string-valued file path to an image
Return:
accuracy of face detection performance in percentage
"""
count = 0
for img in img_path:
if(face_recognition_detector(img)):
count += 1
return (count/len(img_path))*100
human_files_performance = face_recognition_detector_performance(human_files_short)
dog_files_performance = face_recognition_detector_performance(dog_files_short)
print("face detector performace on human files is {:.0f}%".format(human_files_performance))
print("face detector performace on dog files is {:.0f}%".format(dog_files_performance))
Step 2: Detect Dogs¶
In this section, we use a pre-trained ResNet-50 model to detect dogs in images. Our first line of code downloads the ResNet-50 model, along with weights that have been trained on ImageNet, a very large, very popular dataset used for image classification and other vision tasks. ImageNet contains over 10 million URLs, each linking to an image containing an object from one of 1000 categories. Given an image, this pre-trained ResNet-50 model returns a prediction (derived from the available categories in ImageNet) for the object that is contained in the image.
from keras.applications.resnet50 import ResNet50
# define ResNet50 model
ResNet50_model = ResNet50(weights='imagenet')
Pre-process the Data¶
When using TensorFlow as backend, Keras CNNs require a 4D array (which we'll also refer to as a 4D tensor) as input, with shape
$$ (\text{nb_samples}, \text{rows}, \text{columns}, \text{channels}), $$where nb_samples corresponds to the total number of images (or samples), and rows, columns, and channels correspond to the number of rows, columns, and channels for each image, respectively.
The path_to_tensor function below takes a string-valued file path to a color image as input and returns a 4D tensor suitable for supplying to a Keras CNN. The function first loads the image and resizes it to a square image that is $224 \times 224$ pixels. Next, the image is converted to an array, which is then resized to a 4D tensor. In this case, since we are working with color images, each image has three channels. Likewise, since we are processing a single image (or sample), the returned tensor will always have shape
The paths_to_tensor function takes a numpy array of string-valued image paths as input and returns a 4D tensor with shape
Here, nb_samples is the number of samples, or number of images, in the supplied array of image paths. It is best to think of nb_samples as the number of 3D tensors (where each 3D tensor corresponds to a different image) in your dataset!
from keras.preprocessing import image
from tqdm import tqdm
def path_to_tensor(img_path):
"""
Convert a color image to a 4D tensor with shape (1, 224, 224, 3) to supply Keras CNN
Parameters:
img_path: a string-valued file path to an image
Return:
a 4D tensor suitable with shape (1, 224, 224, 3)
"""
# loads RGB image as PIL.Image.Image type
img = image.load_img(img_path, target_size=(224, 224))
# convert PIL.Image.Image type to 3D tensor with shape (224, 224, 3)
x = image.img_to_array(img)
# convert 3D tensor to 4D tensor with shape (1, 224, 224, 3) and return 4D tensor
return np.expand_dims(x, axis=0)
def paths_to_tensor(img_paths):
"""
Convert color images to a 4D tensor with shape of (nb_samples, 224, 224, 3)
Parameters:
img_paths: a numpy array of string-valued image paths
Return:
4D tensor with shape (nb_samples,224,224,3).
"""
list_of_tensors = [path_to_tensor(img_path) for img_path in tqdm(img_paths)]
return np.vstack(list_of_tensors)
Making Predictions with ResNet-50¶
Getting the 4D tensor ready for ResNet-50, and for any other pre-trained model in Keras, requires some additional processing. First, the RGB image is converted to BGR by reordering the channels. All pre-trained models have the additional normalization step that the mean pixel (expressed in RGB as $[103.939, 116.779, 123.68]$ and calculated from all pixels in all images in ImageNet) must be subtracted from every pixel in each image. This is implemented in the imported function preprocess_input. If you're curious, you can check the code for preprocess_input here.
Now that we have a way to format our image for supplying to ResNet-50, we are now ready to use the model to extract the predictions. This is accomplished with the predict method, which returns an array whose $i$-th entry is the model's predicted probability that the image belongs to the $i$-th ImageNet category. This is implemented in the ResNet50_predict_labels function below.
By taking the argmax of the predicted probability vector, we obtain an integer corresponding to the model's predicted object class, which we can identify with an object category through the use of this dictionary.
from keras.applications.resnet50 import preprocess_input, decode_predictions
def ResNet50_predict_labels(img_path):
"""
Use the model to extract the predictions and get the highest predicted object classes
Parameters:
img_path: a string-valued file path to an image
Return:
an integer corresponding to the model's predicted object class
"""
# returns prediction vector for image located at img_path
img = preprocess_input(path_to_tensor(img_path))
return np.argmax(ResNet50_model.predict(img))
Write a Dog Detector¶
While looking at the dictionary, you will notice that the categories corresponding to dogs appear in an uninterrupted sequence and correspond to dictionary keys 151-268, inclusive, to include all categories from 'Chihuahua' to 'Mexican hairless'. Thus, in order to check to see if an image is predicted to contain a dog by the pre-trained ResNet-50 model, we need only check if the ResNet50_predict_labels function above returns a value between 151 and 268 (inclusive).
We use these ideas to complete the dog_detector function below, which returns True if a dog is detected in an image (and False if not).
def dog_detector(img_path):
"""
Detect whether if any dogs in the images
Parameters:
img_path: a string-valued file path to an image
Return:
True: if a dog is detected in an image
False: if a dog is NOT detected in an image
"""
prediction = ResNet50_predict_labels(img_path)
return ((prediction <= 268) & (prediction >= 151))
(IMPLEMENTATION) Assess the Dog Detector¶
Question 3: Use the code cell below to test the performance of your dog_detector function.
- What percentage of the images in
human_files_shorthave a detected dog? - What percentage of the images in
dog_files_shorthave a detected dog?
Answer: 100 percentage of the first 100 images in dog_files_short have been detected with dogs and no image in human_files_short is misclassified as a dog. It shows that the dog detector did well on both of human and dogs images.
def dog_detector_performance(img_path):
"""
Test the performance of the dog_detector algorithm on the images in human_files_short and dog_files_short.
Parameters:
img_path: a string-valued file path to an image
Return:
accuracy of face detection performance in percentage
"""
count = 0
for img in img_path:
if(dog_detector(img)):
count += 1
return (count/len(img_path))*100
human_files_performance = dog_detector_performance(human_files_short)
dog_files_performance = dog_detector_performance(dog_files_short)
print("dog detector performace on human files is {:.0f}%".format(human_files_performance))
print("dog detector performace on dog files is {:.0f}%".format(dog_files_performance))
ResNet50.py¶
from keras.applications.resnet50 import ResNet50
from keras.applications.resnet50 import preprocess_input, decode_predictions
from keras.preprocessing import image
ResNet50_model = ResNet50(weights='imagenet')
class ResNet50:
def __init__(self):
pass
def path_to_tensor(self, img_path):
"""
Convert a color image to a 4D tensor with shape (1, 224, 224, 3) to supply Keras CNN
Parameters:
img_path: a string-valued file path to an image
Return:
a 4D tensor suitable with shape (1, 224, 224, 3)
"""
# loads RGB image as PIL.Image.Image type
img = image.load_img(img_path, target_size=(224, 224))
# convert PIL.Image.Image type to 3D tensor with shape (224, 224, 3)
x = image.img_to_array(img)
# convert 3D tensor to 4D tensor with shape (1, 224, 224, 3) and return 4D tensor
return np.expand_dims(x, axis=0)
def ResNet50_predict_labels(self, img_path):
"""
Use the model to extract the predictions and get the highest predicted object classes
Parameters:
img_path: a string-valued file path to an image
Return:
an integer corresponding to the model's predicted object class
"""
ResNet50_model = self.load_ResNet50_model()
# returns prediction vector for image located at img_path
img = preprocess_input(self.path_to_tensor(img_path))
return np.argmax(ResNet50_model.predict(img))
def dog_detector(self, img_path):
"""
Detect whether if any dogs in the images
Parameters:
img_path: a string-valued file path to an image
Return:
True: if a dog is detected in an image
False: if a dog is NOT detected in an image
"""
prediction = ResNet50_predict_labels(img_path)
return ((prediction <= 268) & (prediction >= 151))
def dog_detector_performance(self, img_path):
"""
Test the performance of the dog_detector algorithm on the images in human_files_short and dog_files_short.
Parameters:
img_path: a string-valued file path to an image
Return:
accuracy of face detection performance in percentage
"""
count = 0
for img in img_path:
if(self.dog_detector(img)):
count += 1
return (count/len(img_path))*100
r = ResNet50()
human_files_performance = r.dog_detector_performance(human_files_short)
dog_files_performance = r.dog_detector_performance(dog_files_short)
print("dog detector performace on human files is {:.0f}%".format(human_files_performance))
print("dog detector performace on dog files is {:.0f}%".format(dog_files_performance))
Step 3: Create a CNN to Classify Dog Breeds (from Scratch)¶
Now that we have functions for detecting humans and dogs in images, we need a way to predict breed from images. In this step, you will create a CNN that classifies dog breeds. You must create your CNN from scratch (so, you can't use transfer learning yet!), and you must attain a test accuracy of at least 1%. In Step 5 of this notebook, you will have the opportunity to use transfer learning to create a CNN that attains greatly improved accuracy.
Be careful with adding too many trainable layers! More parameters means longer training, which means you are more likely to need a GPU to accelerate the training process. Thankfully, Keras provides a handy estimate of the time that each epoch is likely to take; you can extrapolate this estimate to figure out how long it will take for your algorithm to train.
We mention that the task of assigning breed to dogs from images is considered exceptionally challenging. To see why, consider that even a human would have great difficulty in distinguishing between a Brittany and a Welsh Springer Spaniel.
| Brittany | Welsh Springer Spaniel |
|---|---|
 |
 |
It is not difficult to find other dog breed pairs with minimal inter-class variation (for instance, Curly-Coated Retrievers and American Water Spaniels).
| Curly-Coated Retriever | American Water Spaniel |
|---|---|
 |
 |
Likewise, recall that labradors come in yellow, chocolate, and black. Your vision-based algorithm will have to conquer this high intra-class variation to determine how to classify all of these different shades as the same breed.
| Yellow Labrador | Chocolate Labrador | Black Labrador |
|---|---|---|
 |
 |
 |
We also mention that random chance presents an exceptionally low bar: setting aside the fact that the classes are slightly imabalanced, a random guess will provide a correct answer roughly 1 in 133 times, which corresponds to an accuracy of less than 1%.
Remember that the practice is far ahead of the theory in deep learning. Experiment with many different architectures, and trust your intuition. And, of course, have fun!
Pre-process the Data¶
We rescale the images by dividing every pixel in every image by 255.
from PIL import ImageFile
ImageFile.LOAD_TRUNCATED_IMAGES = True
# pre-process the data for Keras
train_tensors = paths_to_tensor(train_files).astype('float32')/255
valid_tensors = paths_to_tensor(valid_files).astype('float32')/255
test_tensors = paths_to_tensor(test_files).astype('float32')/255
(IMPLEMENTATION) Model Architecture¶
Create a CNN to classify dog breed. At the end of your code cell block, summarize the layers of your model by executing the line:
model.summary()
We have imported some Python modules to get you started, but feel free to import as many modules as you need. If you end up getting stuck, here's a hint that specifies a model that trains relatively fast on CPU and attains >1% test accuracy in 5 epochs:
Question 4: Outline the steps you took to get to your final CNN architecture and your reasoning at each step. If you chose to use the hinted architecture above, describe why you think that CNN architecture should work well for the image classification task.
Answer:
epoch = 10; Test accuracy: 3.9474%
| Layers | Output Shape |
|---|---|
| Conv2D | (None, 224, 224, 16) |
| MaxPooling2D | (None, 112, 112, 16) |
| Conv2D | (None, 112, 112, 32) |
| MaxPooling2D | (None, 56, 56, 32) |
| Conv2D | (None, 56, 56, 64) |
| MaxPooling2D | (None, 28, 28, 64) |
| GlobalAveragePooling2D | (None, 64) |
| Dense | (None, 133) |
# epochs = 5 => Test accuracy: 2.8708%%
# epochs = 10 => Test accuracy: 4.3062%
from keras.layers import Conv2D, MaxPooling2D, GlobalAveragePooling2D
from keras.layers import Dropout, Flatten, Dense
from keras.models import Sequential
model = Sequential()
### Define architecture.
model.add(Conv2D(filters=16, kernel_size=2, activation='relu', input_shape=(224, 224, 3)))
model.add(MaxPooling2D(pool_size=2))
model.add(Conv2D(filters=32, kernel_size=2, activation='relu'))
model.add(MaxPooling2D(pool_size=2))
model.add(Conv2D(filters=64, kernel_size=2, activation='relu'))
model.add(MaxPooling2D(pool_size=2))
model.add(GlobalAveragePooling2D())
model.add(Dense(133, activation='softmax'))
model.summary()
### Change kernel_size = 3 ###
# epochs = 5 => Test accuracy: 2.3923%
# epochs = 10 => Test accuracy: 4.4258%
from keras.layers import Conv2D, MaxPooling2D, GlobalAveragePooling2D
from keras.layers import Dropout, Flatten, Dense
from keras.models import Sequential
model = Sequential()
## Define architecture.
model.add(Conv2D(filters=16, kernel_size=3, activation='relu', input_shape=(224, 224, 3)))
model.add(MaxPooling2D(pool_size=2))
model.add(Conv2D(filters=32, kernel_size=3, activation='relu'))
model.add(MaxPooling2D(pool_size=2))
model.add(Conv2D(filters=64, kernel_size=3, activation='relu'))
model.add(MaxPooling2D(pool_size=2))
model.add(GlobalAveragePooling2D())
model.add(Dense(133, activation='softmax'))
model.summary()
##################################
## ##
## Best Model with epoches = 10 ##
## ##
##################################
### Change padding = SAME ###
# epochs = 5 => Test accuracy: 2.1531%
# epochs = 10 => Test accuracy: 3.9474%
from keras.layers import Conv2D, MaxPooling2D, GlobalAveragePooling2D
from keras.layers import Dropout, Flatten, Dense
from keras.models import Sequential
model = Sequential()
## Define architecture.
model.add(Conv2D(filters=16, kernel_size=2, padding='same', activation='relu', input_shape=(224, 224, 3)))
model.add(MaxPooling2D(pool_size=2))
model.add(Conv2D(filters=32, kernel_size=2, padding='same', activation='relu'))
model.add(MaxPooling2D(pool_size=2))
model.add(Conv2D(filters=64, kernel_size=2, padding='same', activation='relu'))
model.add(MaxPooling2D(pool_size=2))
model.add(GlobalAveragePooling2D())
model.add(Dense(133, activation='softmax'))
model.summary()
### Add FC layers ###
# epochs = 5 => Test accuracy: 2.0335%
# epochs = 10 => Test accuracy: 2.9904%
model = Sequential()
## Define architecture.
model.add(Conv2D(filters=16, kernel_size=2, padding='same', activation='relu', input_shape=(224, 224, 3)))
model.add(MaxPooling2D(pool_size=2))
model.add(Conv2D(filters=32, kernel_size=2, padding='same', activation='relu'))
model.add(MaxPooling2D(pool_size=2))
model.add(Conv2D(filters=64, kernel_size=2, padding='same', activation='relu'))
model.add(MaxPooling2D(pool_size=2))
model.add(GlobalAveragePooling2D())
model.add(Dense(64, activation='relu'))
model.add(Dropout(0.3))
model.add(Dense(133, activation='softmax'))
model.summary()
Compile the Model¶
model.compile(optimizer='rmsprop', loss='categorical_crossentropy', metrics=['accuracy'])
(IMPLEMENTATION) Train the Model¶
Train your model in the code cell below. Use model checkpointing to save the model that attains the best validation loss.
You are welcome to augment the training data, but this is not a requirement.
from keras.callbacks import ModelCheckpoint
#the number of epochs that you would like to use to train the model.
epochs = 10
### Do NOT modify the code below this line.
checkpointer = ModelCheckpoint(filepath='saved_models/weights.best.from_scratch.hdf5',
verbose=1, save_best_only=True)
model.fit(train_tensors, train_targets,
validation_data=(valid_tensors, valid_targets),
epochs=epochs, batch_size=20, callbacks=[checkpointer], verbose=1)
Load the Model with the Best Validation Loss¶
model.load_weights('saved_models/weights.best.from_scratch.hdf5')
Test the Model¶
Try out your model on the test dataset of dog images. Ensure that your test accuracy is greater than 1%.
# get index of predicted dog breed for each image in test set
dog_breed_predictions = [np.argmax(model.predict(np.expand_dims(tensor, axis=0))) for tensor in test_tensors]
# report test accuracy
test_accuracy = 100*np.sum(np.array(dog_breed_predictions)==np.argmax(test_targets, axis=1))/len(dog_breed_predictions)
print('Test accuracy: %.4f%%' % test_accuracy)
bottleneck_features = np.load('/data/bottleneck_features/DogVGG16Data.npz')
train_VGG16 = bottleneck_features['train']
valid_VGG16 = bottleneck_features['valid']
test_VGG16 = bottleneck_features['test']
Model Architecture¶
The model uses the the pre-trained VGG-16 model as a fixed feature extractor, where the last convolutional output of VGG-16 is fed as input to our model. We only add a global average pooling layer and a fully connected layer, where the latter contains one node for each dog category and is equipped with a softmax.
VGG16_model = Sequential()
VGG16_model.add(GlobalAveragePooling2D(input_shape=train_VGG16.shape[1:]))
VGG16_model.add(Dense(133, activation='softmax'))
VGG16_model.summary()
Compile the Model¶
VGG16_model.compile(loss='categorical_crossentropy', optimizer='rmsprop', metrics=['accuracy'])
Train the Model¶
checkpointer = ModelCheckpoint(filepath='saved_models/weights.best.VGG16.hdf5',
verbose=1, save_best_only=True)
VGG16_model.fit(train_VGG16, train_targets,
validation_data=(valid_VGG16, valid_targets),
epochs=20, batch_size=20, callbacks=[checkpointer], verbose=1)
Load the Model with the Best Validation Loss¶
VGG16_model.load_weights('saved_models/weights.best.VGG16.hdf5')
Test the Model¶
Now, we can use the CNN to test how well it identifies breed within our test dataset of dog images. We print the test accuracy below.
# get index of predicted dog breed for each image in test set
VGG16_predictions = [np.argmax(VGG16_model.predict(np.expand_dims(feature, axis=0))) for feature in test_VGG16]
# report test accuracy
test_accuracy = 100*np.sum(np.array(VGG16_predictions)==np.argmax(test_targets, axis=1))/len(VGG16_predictions)
print('Test accuracy: %.4f%%' % test_accuracy)
Predict Dog Breed with the Model¶
from extract_bottleneck_features import *
def VGG16_predict_breed(img_path):
"""
Use VGG16 model to predicte dog breed
Parameters:
img_path: a string-valued file path to an image
Return:
return dog breed that is predicted by the model
"""
# extract bottleneck features
bottleneck_feature = extract_VGG16(path_to_tensor(img_path))
# obtain predicted vector
predicted_vector = VGG16_model.predict(bottleneck_feature)
# return dog breed that is predicted by the model
return dog_names[np.argmax(predicted_vector)]
Step 5: Create a CNN to Classify Dog Breeds (using Transfer Learning)¶
You will now use transfer learning to create a CNN that can identify dog breed from images. Your CNN must attain at least 60% accuracy on the test set.
In Step 4, we used transfer learning to create a CNN using VGG-16 bottleneck features. In this section, you must use the bottleneck features from a different pre-trained model. To make things easier for you, we have pre-computed the features for all of the networks that are currently available in Keras. These are already in the workspace, at /data/bottleneck_features. If you wish to download them on a different machine, they can be found at:
- VGG-19 bottleneck features
- ResNet-50 bottleneck features
- Inception bottleneck features
- Xception bottleneck features
The files are encoded as such:
Dog{network}Data.npz
where {network}, in the above filename, can be one of VGG19, Resnet50, InceptionV3, or Xception.
The above architectures are downloaded and stored for you in the /data/bottleneck_features/ folder.
This means the following will be in the /data/bottleneck_features/ folder:
DogVGG19Data.npz
DogResnet50Data.npz
DogInceptionV3Data.npz
DogXceptionData.npz
(IMPLEMENTATION) Obtain Bottleneck Features¶
In the code block below, extract the bottleneck features corresponding to the train, test, and validation sets by running the following:
bottleneck_features = np.load('/data/bottleneck_features/Dog{network}Data.npz')
train_{network} = bottleneck_features['train']
valid_{network} = bottleneck_features['valid']
test_{network} = bottleneck_features['test']### Obtain bottleneck features from InceptionV3 pre-trained CNN.
bottleneck_features = np.load('/data/bottleneck_features/DogInceptionV3Data.npz')
train_InceptionV3 = bottleneck_features['train']
valid_InceptionV3 = bottleneck_features['valid']
test_InceptionV3 = bottleneck_features['test']
(IMPLEMENTATION) Model Architecture¶
Create a CNN to classify dog breed. At the end of your code cell block, summarize the layers of your model by executing the line:
<your model's name>.summary()
Question 5: Outline the steps you took to get to your final CNN architecture and your reasoning at each step. Describe why you think the architecture is suitable for the current problem.
Answer:
Step 1: Build different architectures¶
Architecture 1:
| Layers | Output Shape |
|---|---|
| GlobalAveragePooling2D | (None, 2048) |
| Dense | (None, 133) |
Architecture 2: add hidden layer in classifier
| Layers | Output Shape |
|---|---|
| GlobalAveragePooling2D | (None, 2048) |
| Dense | (None, 128) |
| Dense | (None, 133) |
Test accuracy:
| Architecture 1 | Architecture 2 |
|---|---|
| 80.0239% (epoch = 5, batch_size = 20) | 78.4689% (epoch = 5, batch_size = 20) |
| 80.6220% (epoch = 10, batch_size = 20) | 80.8612% (epoch = 10, batch_size = 20) |
| 80.5024% (epoch = 20, batch_size = 20) | 80.6220% (epoch = 20, batch_size = 20) |
| 77.6316% (epoch = 25, batch_size = 20) | 80.0239% (epoch = 25, batch_size = 20) |
Result1: It shows that Architecture 1 and Architecture 2 have better performance as epoch is 10. But can't tell which architecture has better performance at this point. Will keep both and try different parameters.
Step 2: Test on augmented images¶
Test accuracy:
| Architecture 1 | Architecture 2 |
|---|---|
| 76.9139% (epoch = 5, batch_size = 20) | 79.9043% (epoch = 5, batch_size = 20) |
| 80.6220% (epoch = 10, batch_size = 20) | 79.5455% (epoch = 10, batch_size = 20) |
| 80.7416% (epoch = 15, batch_size = 20) | N/A (epoch = 20, batch_size = 20) |
Result2: It shows that Architecture 1 have better performance on test augmented images than Architecture 1 as epoch is more than 10. Therefore, I will choose Architecture 1 as model to tune other hyperparameters.
Step 3: Try different batch sizes¶
Test accuracy:
| Architecture 1 |
|---|
| 80.6220% (epoch = 10, batch_size = 20) |
| 80.3828% (epoch = 10, batch_size = 30) |
| 82.0574% (epoch = 10, batch_size = 40) |
| 81.5789% (epoch = 10, batch_size = 60) |
| 81.4593% (epoch = 10, batch_size = 120) |
Result3: It shows that Architecture 1 have better performance as batch_size is 40.
Step 4: Test on augmented images with epoch=10, batch_size=40 by model of Architecture 1¶
Test accuracy:
| w/o augmented images | w/ augmented images |
|---|---|
| 82.0574% | 82.0574% |
Result4: It turns out the model has the same performance on either original images or augmented images. Because it is hard to expect what kind of images users will used to test, I will take the weights which is tested on augmented images as final model weights.
###################################################
## ##
## Best Model with epoches = 10, batch_size = 40 ##
## ##
###################################################
### test w/o augmented image ###
# epoch = 5, batch_size = 20, Test accuracy:80.0239%
# epoch = 10, batch_size = 20, Test accuracy: 80.6220%
# epoch = 20, batch_size = 20, Test accuracy: 80.5024%
# epoch = 25, batch_size = 20, Test accuracy: 77.6316%
# epoch = 10, batch_size = 30, Test accuracy: 80.3828%
# epoch = 10, batch_size = 40, Test accuracy: 82.0574% => best
# epoch = 10, batch_size = 60, Test accuracy: 81.5789%
# epoch = 10, batch_size = 120, Test accuracy: 81.4593%
### test w/ augmented images ###
# epoch = 5, batch_size = 20, Test accuracy:76.9139%
# epoch = 10, batch_size = 20, Test accuracy: 80.6220%
# epoch = 15, batch_size = 20, Test accuracy: 80.7416%
# epoch = 10, batch_size = 40, Test accuracy: 82.0574% => best => final save weights for model
from keras.layers import Conv2D, MaxPooling2D, GlobalAveragePooling2D
from keras.layers import Dropout, Flatten, Dense
from keras.models import Sequential
### Define architecture.
InceptionV3_model = Sequential()
InceptionV3_model.add(GlobalAveragePooling2D(input_shape=train_InceptionV3.shape[1:]))
InceptionV3_model.add(Dense(133, activation='softmax'))
InceptionV3_model.summary()
#####Add more fc layers#####
### no augmented image ###
# epoch = 5, batch_size = 20, Test accuracy: 78.4689%
# epoch = 10, batch_size = 20, Test accuracy: 80.8612%
# epoch = 20, batch_size = 20, Test accuracy: 80.6220%
# epoch = 25, batch_size = 20, Test accuracy: 80.0239%
### test w/ augmented images ###
# epoch = 5, batch_size = 20, Test accuracy: 79.9043%
# epoch = 10, batch_size = 20, Test accuracy: 79.5455%
InceptionV3_model = Sequential()
InceptionV3_model.add(GlobalAveragePooling2D(input_shape=train_InceptionV3.shape[1:]))
InceptionV3_model.add(Dense(128, activation='relu'))
InceptionV3_model.add(Dropout(0.5))
InceptionV3_model.add(Dense(133, activation='softmax'))
InceptionV3_model.summary()
(IMPLEMENTATION) Compile the Model¶
### Compile the model.
InceptionV3_model.compile(loss='categorical_crossentropy', optimizer='rmsprop', metrics=['accuracy'])
Create and configure augmented image generator¶
from keras.preprocessing.image import ImageDataGenerator
datagen_train = ImageDataGenerator(
rotation_range=40,
width_shift_range=0.2,
height_shift_range=0.2,
shear_range=0.2,
zoom_range=0.2,
horizontal_flip=True,
fill_mode='nearest')
Visualize Oroginal and Augmented Images¶
x_train_subset = train_tensors[:12]
fig = plt.figure(figsize=(20,2))
for i in range(0, len(x_train_subset)):
ax = fig.add_subplot(1, 12, i+1)
ax.imshow(x_train_subset[i])
fig.suptitle("Subset of Original Training Images", fontsize=15)
plt.show()
fig = plt.figure(figsize=(20,2))
for x_batch in datagen_train.flow(x_train_subset, batch_size=12):
for i in range(0, len(x_train_subset)):
ax = fig.add_subplot(1, 12, i+1)
ax.imshow(x_batch[i])
fig.suptitle("Augmented Images", fontsize=15)
plt.show()
break;
(IMPLEMENTATION) Train the Model¶
Train your model in the code cell below. Use model checkpointing to save the model that attains the best validation loss.
You are welcome to augment the training data, but this is not a requirement.
### Train the model.
checkpointer = ModelCheckpoint(filepath='saved_models/weights.best.InceptionV3.hdf5',
verbose=1, save_best_only=True)
InceptionV3_model.fit(train_InceptionV3, train_targets,
validation_data=(valid_InceptionV3, valid_targets),
epochs=10, batch_size=30, callbacks=[checkpointer], verbose=1)
Train the Model with augmented images¶
batch_size=40
checkpointer = ModelCheckpoint(filepath='saved_models/weights.best.InceptionV3.hdf5',
verbose=1, save_best_only=True)
InceptionV3_model.fit_generator(datagen_train.flow(train_InceptionV3, train_targets, batch_size=batch_size),
steps_per_epoch= train_InceptionV3.shape[0]//batch_size,
validation_data=(valid_InceptionV3, valid_targets),
epochs=10,callbacks=[checkpointer], verbose=1)
(IMPLEMENTATION) Load the Model with the Best Validation Loss¶
### Load the model weights with the best validation loss.
InceptionV3_model.load_weights('saved_models/weights.best.InceptionV3.hdf5')
(IMPLEMENTATION) Test the Model¶
Try out your model on the test dataset of dog images. Ensure that your test accuracy is greater than 60%.
### Calculate classification accuracy on the test dataset.
# get index of predicted dog breed for each image in test set
InceptionV3_predictions = [np.argmax(InceptionV3_model.predict(np.expand_dims(feature, axis=0))) for feature in test_InceptionV3]
# report test accuracy
test_accuracy = 100*np.sum(np.array(InceptionV3_predictions)==np.argmax(test_targets, axis=1))/len(InceptionV3_predictions)
print('Test accuracy: %.4f%%' % test_accuracy)
(IMPLEMENTATION) Predict Dog Breed with the Model¶
Write a function that takes an image path as input and returns the dog breed (Affenpinscher, Afghan_hound, etc) that is predicted by your model.
Similar to the analogous function in Step 5, your function should have three steps:
- Extract the bottleneck features corresponding to the chosen CNN model.
- Supply the bottleneck features as input to the model to return the predicted vector. Note that the argmax of this prediction vector gives the index of the predicted dog breed.
- Use the
dog_namesarray defined in Step 0 of this notebook to return the corresponding breed.
The functions to extract the bottleneck features can be found in extract_bottleneck_features.py, and they have been imported in an earlier code cell. To obtain the bottleneck features corresponding to your chosen CNN architecture, you need to use the function
extract_{network}
where {network}, in the above filename, should be one of VGG19, Resnet50, InceptionV3, or Xception.
from extract_bottleneck_features import *
def InceptionV3_predict_breed(img_path):
"""
Use InceptionV3 model to predicte dog breed
Parameters:
img_path: a string-valued file path to an image
Return:
return dog breed that is predicted by the model
"""
# extract bottleneck features
bottleneck_feature = extract_InceptionV3(path_to_tensor(img_path))
# obtain predicted vector
predicted_vector = InceptionV3_model.predict(bottleneck_feature)
# return dog breed that is predicted by the model
return dog_names[np.argmax(predicted_vector)]
from extract_bottleneck_features import *
def InceptionV3_predict_top5_breeds(img_path):
"""
Use InceptionV3 model to predicte dog breed and return top5 predicted breeds
Parameters:
img_path: a string-valued file path to an image
Return:
top5_breeds_idx: top5 dog breed idx that is predicted by the model
top5_breeds_prob: top5 dog breed probabilities that are corresponding to top5 dog breed idx
"""
# extract bottleneck features
bottleneck_feature = extract_InceptionV3(path_to_tensor(img_path))
# obtain predicted vector
predicted_vector = InceptionV3_model.predict(bottleneck_feature)
# return dog breed that is predicted by the model
# np.argsort(-predicted_vector) return matrix shape (1, 133)
top5_breeds_idx= np.argsort(-predicted_vector)[0][:5] # return top5 probability class idx
top5_breeds_prob= [predicted_vector[0][idx] for idx in top5_breeds_idx]# return top5 probability
return top5_breeds_idx, top5_breeds_prob
Step 6: Write your Algorithm¶
Write an algorithm that accepts a file path to an image and first determines whether the image contains a human, dog, or neither. Then,
- if a dog is detected in the image, return the predicted breed.
- if a human is detected in the image, return the resembling dog breed.
- if neither is detected in the image, provide output that indicates an error.
You are welcome to write your own functions for detecting humans and dogs in images, but feel free to use the face_detector and dog_detector functions developed above. You are required to use your CNN from Step 5 to predict dog breed.
Some sample output for our algorithm is provided below, but feel free to design your own user experience!
(IMPLEMENTATION) Write your Algorithm¶
Snapchat-like filter¶
# load dog_filter images
dog_left_ear = cv2.imread('filters/dog_left_ear.png', cv2.IMREAD_UNCHANGED) # rgba
dog_right_ear = cv2.imread('filters/dog_right_ear.png',cv2.IMREAD_UNCHANGED)
dog_nose = cv2.imread('filters/dog_nose.png', cv2.IMREAD_UNCHANGED)
dog_tongue = cv2.imread('filters/dog_tongue.png', cv2.IMREAD_UNCHANGED)
def resize_filter(img, w):
"""
Adjust the size of filter images according to the size of the detected face on the test image
Parameters:
img: images that has read by OpenCV (BGR image)
w: width of bounding box for the detected face
Return:
img: resized image
"""
new_w = int(w*0.3) # resize the image to 0.3 scale of the width of the face detector
new_h = int(img.shape[0]/img.shape[1]*new_w) # keep aspect ratio
img = cv2.resize(img, (new_w, new_h))
return img
def put_dog_left_ear(dog_left_ear_img, human_img, x, y, w, h):
"""
Overlay dog_left_ear filter image to the detected faces on human images
Parameters:
dog_left_ear_img: filter image with dog left ear
human_img: input image for the classifier
x, y : coordinates of the bounding box for the detected face at the top right corner
w: width of bounding box for the detected face
h: height of bounding box for the detected face
Return:
human_img: human image with dog left ear filter
"""
dog_leftEar_img = resize_filter(dog_left_ear_img, w)
#overlay range
yo = dog_leftEar_img.shape[0]
xo = dog_leftEar_img.shape[1]
#loop every pixel
for j in range(yo):
for i in range(xo):
for k in range(3):
alpha = float(dog_leftEar_img[j][i][3]/255.0) # read the alpha channel
#human_img[x + i][y + j][k] = dog_ear_img[i][j][k] # with black background if w/o alpha
human_img[y+j][x+i][k] = alpha*dog_leftEar_img[j][i][k]+(1-alpha)*human_img[y+j][x+i][k]
return human_img
def put_dog_right_ear(dog_right_ear_img, human_img, x, y, w, h):
"""
Overlay dog_right_ear filter image to the detected faces on human images
Parameters:
dog_right_ear_img: filter image with dog right ear
human_img: input image for the classifier
x, y : coordinates of the bounding box for the detected face at the top right corner
w: width of bounding box for the detected face
h: height of bounding box for the detected face
Return:
human_img: human image with dog right ear filter
"""
dog_rightEar_img = resize_filter(dog_right_ear_img, w)
#overlay range
yo = dog_rightEar_img.shape[0]
xo = dog_rightEar_img.shape[1]
#loop every pixel
for j in range(yo):
for i in range(xo):
for k in range(3):
alpha = float(dog_rightEar_img[j][i][3]/255.0) # read the alpha channel
human_img[y+j][x+w-xo+i][k] = alpha*dog_rightEar_img[j][i][k]+(1-alpha)*human_img[y+j][x+w-xo+i][k]
return human_img
def put_dog_nose(dog_nose_img, human_img, x, y, w, h):
"""
Overlay dog_nose filter image to the detected faces on human images
Parameters:
dog_nose_img: filter image with dog nose
human_img: input image for the classifier
x, y : coordinates of the bounding box for the detected face at the top right corner
w: width of bounding box for the detected face
h: height of bounding box for the detected face
Return:
human_img: human image with dog nose filter
"""
dog_Nose_img = resize_filter(dog_nose_img, w)
#overlay range
yo = dog_Nose_img.shape[0]
xo = dog_Nose_img.shape[1]
#loop every pixel
for j in range(yo):
for i in range(xo):
for k in range(3):
alpha = float(dog_Nose_img[j][i][3]/255.0) # read the alpha channel
human_img[y+int(h/2)+j][x+int((w-xo)/2)+i][k] = alpha*dog_Nose_img[j][i][k]+\
(1-alpha)*human_img[y+int(h/2)+j][x+int((w-xo)/2)+i][k]
return human_img
def put_dog_tongue(dog_tongue_img, human_img, x, y, w, h):
"""
Overlay dog_tongue filter image to the detected faces on human images
Parameters:
dog_tongue_img: filter image with dog tongue
human_img: input image for the classifier
x, y : coordinates of the bounding box for the detected face at the top right corner
w: width of bounding box for the detected face
h: height of bounding box for the detected face
Return:
human_img: human image with dog tongue filter
"""
dog_Tongue_img = resize_filter(dog_tongue_img, w)
#overlay range
yo = dog_Tongue_img.shape[0]
xo = dog_Tongue_img.shape[1]
#loop every pixel
for j in range(yo):
for i in range(xo):
for k in range(3):
alpha = float(dog_Tongue_img[j][i][3]/255.0) # read the alpha channel
human_img[y+h-yo+j][x+int((w-xo)/2)+i][k] = alpha*dog_Tongue_img[j][i][k]+\
(1-alpha)*human_img[y+h-yo+j][x+int((w-xo)/2)+i][k]
return human_img
def apply_snapchat_filter(img_path):
"""
Overlay dog filter images(left_ear, right_ear, nose, tongue) to the detected faces on human images
Parameters:
img_path: a string-valued file path to an image
Return:
None
"""
# extract pre-trained face detector
face_cascade = cv2.CascadeClassifier('haarcascades/haarcascade_frontalface_alt.xml')
# load color (BGR) image
img = cv2.imread(img_path)
# convert BGR image to grayscale
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
# find faces in image
face = face_cascade.detectMultiScale(gray)
# add overlayer
for (x,y,w,h) in face:
img = put_dog_left_ear(dog_left_ear, img, x, y, w, h)
img = put_dog_right_ear(dog_right_ear, img, x, y, w, h)
img = put_dog_nose(dog_nose, img, x, y, w, h)
img = put_dog_tongue(dog_tongue, img, x, y, w, h)
cv_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
fig = plt.figure(figsize=(10, 5))
ax = fig.add_subplot(1, 2, 2, xticks=[], yticks=[])
plt.imshow(cv_rgb)
Dog_filters.py¶
import cv2
import matplotlib.pyplot as plt
dog_left_ear = cv2.imread('filters/dog_left_ear.png', cv2.IMREAD_UNCHANGED) # rgba
dog_right_ear = cv2.imread('filters/dog_right_ear.png',cv2.IMREAD_UNCHANGED)
dog_nose = cv2.imread('filters/dog_nose.png', cv2.IMREAD_UNCHANGED)
dog_tongue = cv2.imread('filters/dog_tongue.png', cv2.IMREAD_UNCHANGED)
dog_filters = [dog_left_ear, dog_right_ear, dog_nose, dog_tongue]
class Dog_filters:
def __init__(self, face_img_path, scale=0.3):
self.face_img_path = face_img_path
self.face_img = cv2.imread(self.face_img_path) # load color (BGR) image
self.scale = scale
self.resize_filters=[]
def resize_filter(self, w):
"""
Adjust the size of filter images according to the size of the detected face on the test image
Parameters:
img: images that has read by OpenCV (BGR image)
w: width of bounding box for the detected face
Return:
img: resized image
"""
for dog_filter in dog_filters:
new_w = int(w*0.3) # resize the image to 0.3 scale of the width of the face detector
new_h = int(dog_filter.shape[0]/dog_filter.shape[1]*new_w) # keep aspect ratio
resize_filter = cv2.resize(dog_filter, (new_w, new_h))
self.resize_filters.append(resize_filter)
def put_dog_left_ear(self, x, y):
"""
Overlay dog_left_ear filter image to the detected faces on human images
Parameters:
dog_left_ear_img: filter image with dog left ear
human_img: input image for the classifier
x, y : coordinates of the bounding box for the detected face at the top right corner
w: width of bounding box for the detected face
h: height of bounding box for the detected face
Return:
human_img: human image with dog left ear filter
"""
dog_left_ear_filter = self.resize_filters[0]
#overlay range
yo = dog_left_ear_filter.shape[0]
xo = dog_left_ear_filter.shape[1]
#loop every pixel
for j in range(yo):
for i in range(xo):
for k in range(3):
alpha = float(dog_left_ear_filter[j][i][3]/255.0) # read the alpha channel
#self.face_img[x + i][y + j][k] = dog_left_ear_filter[i][j][k] # with black background if w/o alpha
self.face_img[y+j][x+i][k] = alpha*dog_left_ear_filter[j][i][k]+(1-alpha)*self.face_img[y+j][x+i][k]
def put_dog_right_ear(self, x, y, w, h):
"""
Overlay dog_right_ear filter image to the detected faces on human images
Parameters:
dog_right_ear_img: filter image with dog right ear
human_img: input image for the classifier
x, y : coordinates of the bounding box for the detected face at the top right corner
w: width of bounding box for the detected face
h: height of bounding box for the detected face
Return:
human_img: human image with dog right ear filter
"""
dog_right_ear_filter = self.resize_filters[1]
#overlay range
yo = dog_right_ear_filter.shape[0]
xo = dog_right_ear_filter.shape[1]
#loop every pixel
for j in range(yo):
for i in range(xo):
for k in range(3):
alpha = float(dog_right_ear_filter[j][i][3]/255.0) # read the alpha channel
self.face_img[y+j][x+w-xo+i][k] = alpha*dog_right_ear_filter[j][i][k]+(1-alpha)*self.face_img[y+j][x+w-xo+i][k]
def put_dog_nose(self, x, y, w, h):
"""
Overlay dog_nose filter image to the detected faces on human images
Parameters:
dog_nose_img: filter image with dog nose
human_img: input image for the classifier
x, y : coordinates of the bounding box for the detected face at the top right corner
w: width of bounding box for the detected face
h: height of bounding box for the detected face
Return:
human_img: human image with dog nose filter
"""
dog_nose_filter = self.resize_filters[2]
#overlay range
yo = dog_nose_filter.shape[0]
xo = dog_nose_filter.shape[1]
#loop every pixel
for j in range(yo):
for i in range(xo):
for k in range(3):
alpha = float(dog_nose_filter[j][i][3]/255.0) # read the alpha channel
self.face_img[y+int(h/2)+j][x+int((w-xo)/2)+i][k] = alpha*dog_nose_filter[j][i][k]+\
(1-alpha)*self.face_img[y+int(h/2)+j][x+int((w-xo)/2)+i][k]
def put_dog_tongue(self, x, y, w, h):
"""
Overlay dog_tongue filter image to the detected faces on human images
Parameters:
dog_tongue_img: filter image with dog tongue
human_img: input image for the classifier
x, y : coordinates of the bounding box for the detected face at the top right corner
w: width of bounding box for the detected face
h: height of bounding box for the detected face
Return:
human_img: human image with dog tongue filter
"""
dog_nose_filter = self.resize_filters[3]
#overlay range
yo = dog_nose_filter.shape[0]
xo = dog_nose_filter.shape[1]
#loop every pixel
for j in range(yo):
for i in range(xo):
for k in range(3):
alpha = float(dog_nose_filter[j][i][3]/255.0) # read the alpha channel
self.face_img[y+h-yo+j][x+int((w-xo)/2)+i][k] = alpha*dog_nose_filter[j][i][k]+\
(1-alpha)*self.face_img[y+h-yo+j][x+int((w-xo)/2)+i][k]
def apply_snapchat_filter(self):
"""
Overlay dog filter images(left_ear, right_ear, nose, tongue) to the detected faces on human images
Parameters:
img_path: a string-valued file path to an image
Return:
None
"""
# extract pre-trained face detector
face_cascade = cv2.CascadeClassifier('haarcascades/haarcascade_frontalface_alt.xml')
# convert BGR image to grayscale
gray = cv2.cvtColor(self.face_img, cv2.COLOR_BGR2GRAY)
# find faces in image
face = face_cascade.detectMultiScale(gray)
# add overlayer
for (x,y,w,h) in face:
self.resize_filter(w)
self.put_dog_left_ear(x, y)
self.put_dog_right_ear(x, y, w, h)
self.put_dog_nose(x, y, w, h)
self.put_dog_tongue(x, y, w, h)
cv_rgb = cv2.cvtColor(self.face_img, cv2.COLOR_BGR2RGB)
fig = plt.figure(figsize=(10, 5))
ax = fig.add_subplot(1, 2, 2, xticks=[], yticks=[])
plt.imshow(cv_rgb)
Visualize snapchat-like filter on human face¶
import cv2
import matplotlib.pyplot as plt
%matplotlib inline
df = Dog_filters(human_files[40])
df.apply_snapchat_filter()
def isMix_mutt(top5_breeds_prob, top5_labels):
"""
Decide whether if the detected dog/human is resemble to mix mutt dog breeds or not.
If the difference of top 1 probability and top 2 probability is less than 0.1, it will be classified as
mix mutt.
Parameters:
top5_breeds_prob: top5 dog breed probabilities that is predicted by the model
top5_labels: top5 dog breed labels that are corresponding to top5 dog breed idx
Return:
None
"""
if (top5_breeds_prob[0] - top5_breeds_prob[1] <= 0.1):
print("You look like", top5_labels[0], '&', top5_labels[1] ,"mixed breed dog!")
else:
print("You look like", top5_labels[0], "!")
def show_top5_result(img_path):
"""
Visualize top 5 dog breeds that are predicted by the model via bar chart
Parameters:
None
Return:
None
"""
top5_breeds_idx, top5_breeds_prob = InceptionV3_predict_top5_breeds(img_path)
top5_labels = [dog_names[idx] for idx in top5_breeds_idx]
isMix_mutt(top5_breeds_prob, top5_labels)
#plot top5 result
plt.subplot(1,2,1)
plt.barh(top5_labels, top5_breeds_prob)
plt.xlabel('probability')
plt.xlim(0, 1.0)
plt.title('dog breeds classifer')
plt.show()
InceptionV3.py¶
from dog_names import dog_names
from keras.models import Sequential
from keras.layers import GlobalAveragePooling2D, Dense
from keras.preprocessing import image
from extract_bottleneck_features import *
import numpy as np
import matplotlib.pyplot as plt
class InceptionV3:
def __init__(self, img_path, InceptionV3_model=None, input_shape=(5,5,2048)):
self.img_path = img_path
if InceptionV3_model is None:
model = Sequential()
model.add(GlobalAveragePooling2D(input_shape = input_shape))
model.add(Dense(133, activation='softmax'))
model.load_weights('saved_models/weights.best.InceptionV3.hdf5')
InceptionV3_model = model
self.InceptionV3_model = InceptionV3_model
self.input_shape = input_shape
def path_to_tensor(self):
"""
Convert a color image to a 4D tensor with shape (1, 224, 224, 3) to supply Keras CNN
Parameters:
img_path: a string-valued file path to an image
Return:
a 4D tensor suitable with shape (1, 224, 224, 3)
"""
# loads RGB image as PIL.Image.Image type
img = image.load_img(self.img_path, target_size=(224, 224))
# convert PIL.Image.Image type to 3D tensor with shape (224, 224, 3)
x = image.img_to_array(img)
# convert 3D tensor to 4D tensor with shape (1, 224, 224, 3) and return 4D tensor
return np.expand_dims(x, axis=0)
def predict_top5_breeds(self):
"""
Use InceptionV3 model to predicte dog breed and return top5 predicted breeds
Parameters:
None
Return:
top5_breeds_idx: top5 dog breed idx that is predicted by the model
top5_breeds_prob: top5 dog breed probabilities that are corresponding to top5 dog breed idx
"""
# extract bottleneck features
bottleneck_feature = extract_InceptionV3(self.path_to_tensor())
# obtain predicted vector
predicted_vector = self.InceptionV3_model.predict(bottleneck_feature)
# return dog breed that is predicted by the model
# np.argsort(-predicted_vector) return matrix shape (1, 133)
top5_breeds_idx= np.argsort(-predicted_vector)[0][:5] # return top5 probability class idx
top5_breeds_prob= [predicted_vector[0][idx] for idx in top5_breeds_idx]# return top5 probability
return top5_breeds_idx, top5_breeds_prob
def isMix_mutt(self, top5_breeds_prob, top5_labels):
"""
Decide whether if the detected dog/human is resemble to mix mutt dog breeds or not.
If the difference of top 1 probability and top 2 probability is less than 0.1, it will be classified as
mix mutt.
Parameters:
top5_breeds_prob: top5 dog breed probabilities that is predicted by the model
top5_labels: top5 dog breed labels that are corresponding to top5 dog breed idx
Return:
None
"""
if (top5_breeds_prob[0] - top5_breeds_prob[1] <= 0.1):
print("You look like", top5_labels[0], '&', top5_labels[1] ,"mixed breed dog!")
else:
print("You look like", top5_labels[0], "!")
def show_top5_result(self):
"""
Visualize top 5 dog breeds that are predicted by the model via bar chart
Parameters:
None
Return:
None
"""
top5_breeds_idx, top5_breeds_prob = self.predict_top5_breeds()
top5_labels = [dog_names[idx] for idx in top5_breeds_idx]
self.isMix_mutt(top5_breeds_prob, top5_labels)
#plot top5 result
plt.subplot(1,2,1)
plt.barh(top5_labels, top5_breeds_prob)
plt.xlabel('probability')
plt.xlim(0, 1.0)
plt.title('dog breeds classifer')
plt.show()
util/helper functions¶
def crop_detected_faces(BGR_img, faces):
"""
Crop faces on the test human image
Parameters:
BGR_img: a color (BGR) image has been read by OpenCV
faces: number of faces that have been detected in an image
Return:
cropped_imgs: cropped images with detected faces
"""
cropped_imgs = []
offset = 25
height, width = BGR_img.shape[:2]
for i in range(len(faces)):
x,y,w,h = faces[i]
cropped_img = BGR_img[y-offset : y+h+offset, x-offset : x+w+offset]
if(y-offset < 0):
cropped_img = BGR_img[0 : y+h+offset, x-offset : x+w+offset]
if(x-offset < 0):
cropped_img = BGR_img[y-offset : y+h+offset, 0 : x+w+offset]
if(y+h+offset > height):
cropped_img = BGR_img[y-offset : height, 0 : x+w+offset]
if(x+w+offset > width):
cropped_img = BGR_img[y-offset : y+h+offset, 0 : width]
cropped_imgs.append(cropped_img)
return cropped_imgs
import os
def save_cropped_imgs(num, cropped_imgs):
"""
Save cropped images to temp directory for later use
Parameters:
cropped_imgs: cropped images with detected faces
num: number of cropped images from a test image
Return:
cropped_img_path: a string-valued file path to a cropped image
"""
dirName = 'tempDir'
if not os.path.exists(dirName):
os.mkdir(dirName)
cropped_img_path = dirName + '/img' + str(num) + '.jpg'
cv2.imwrite(cropped_img_path, cropped_imgs[num])
return cropped_img_path
import glob
def delete_cropped_images():
"""
Delete all files in temp directory
Parameters:
None
Return:
None
"""
files = glob.glob('tempDir/*')
for f in files:
os.remove(f)
def show_test_image(img_path):
"""
Display the test image on predicted result
Parameters:
img_path: a string-valued file path to an image
Return:
None
"""
fig = plt.figure(figsize=(10, 5))
ax = fig.add_subplot(1, 2, 2, xticks=[], yticks=[])
visualize_img(img_path, ax)
Predicted functions¶
def predict_breed(img_path):
"""
Predict which dog breeds the image is resemble.
if any face is detected, it will show "Hello, human!" and predict dog breeds result.
if any dog is detected, it will show "Hello, dog!" and predict dog breeds result.
if no dog and human are detected, it will show "No human. No dog."
Parameters:
img_path: a string-valued file path to an image
Return:
None
"""
faces = OpenCV_face_detector(img_path)
dogs = dog_detector(img_path)
if(faces):
print('Hello, human!')
apply_snapchat_filter(img_path)
show_top5_result(img_path)
if(dogs):
print('Hello, dog!')
show_test_image(img_path)
show_top5_result(img_path)
if(faces == 0 and dogs == 0):
print('No human. No dog.')
show_test_image(img_path)
def detect_face_on_cropped_imgs(cropped_imgs):
"""
Detect whether if any faces in an image
Parameters:
img_path: a string-valued file path to an image
Return:
True: if face is detected in image stored at img_path
False: if face is NOT detected in image stored at img_path
"""
for i in range(len(cropped_imgs)):
cropped_img_path = save_cropped_imgs(i, cropped_imgs)
predict_breed_for_human_only(cropped_img_path)
def improved_predict_breed(img_path):
"""
Predict which dog breeds the image is resemble.
if dog and face are detected on the same image
- predict dog breeds on the detected dog
- crop the detected face area on the image and predict dog breeds on the detected face
if more than one detected faces on the same image
- crop the detected face area on the image and predict dog breeds on the individual detected face
if only one face is detected on the same image
- predict dog breeds on the detected face
if no dog and human are detected
- print "No human. No dog."
Parameters:
img_path: a string-valued file path to an image
Return:
None
"""
faces, BGR_img = Improved_OpenCV_face_detector(img_path)
dogs = dog_detector(img_path)
#if dog and human in the same image, model predicts dog breeds will always based on the dog
#so we have to cropped the human image from the dog
if(dogs != 0):
print('Hello, dog!')
show_test_image(img_path)
show_top5_result(img_path)
if(len(faces) > 0):
cropped_imgs = crop_detected_faces(BGR_img, faces)
detect_face_on_cropped_imgs(cropped_imgs)
delete_cropped_images()
#if more than one people in the same image, model predicts dog breeds will always show one result
#so we have to crop the human image to individuals
else:
if(len(faces) > 1):
cropped_imgs = crop_detected_faces(BGR_img, faces)
detect_face_on_cropped_imgs(cropped_imgs)
delete_cropped_images()
elif(len(faces) == 1):
print('Hello, human!')
apply_snapchat_filter(img_path)
show_top5_result(img_path)
else:
print('No human. No dog.')
show_test_image(img_path)
def predict_breed_for_human_only(img_path):
"""
Predict which dog breeds the cropped image is resemble.
cropped image includes only one detected face, so no need to run the whole process in improved_predict_breed function
Parameters:
img_path: a string-valued file path to an image
Return:
None
"""
faces, BGR_img = Improved_OpenCV_face_detector(img_path)
print('Hello, human!')
apply_snapchat_filter(img_path)
show_top5_result(img_path)
Predict_breeds.py¶
import dog_names as dn
from dog_filters import Dog_filters as df
from InceptionV3 import InceptionV3 as iV3
from ResNet50 import ResNet50 as rn50
import util as u
import numpy as np
import cv2
class Predict_breeds:
def __init__(self, img_path):
self.img_path = img_path
def Improved_OpenCV_face_detector(self):
"""
Detect whether if any faces in an image
Parameters:
img_path: a string-valued file path to an image
Return:
faces: number of faces that have been detected in an image
img: a color (BGR) image has been read by OpenCV
"""
face_cascade = cv2.CascadeClassifier('haarcascades/haarcascade_frontalface_alt.xml')
img = cv2.imread(self.img_path) #BGR image
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
faces = face_cascade.detectMultiScale(gray)
return faces, img
def predict_breed_for_human_only(self):
"""
Predict which dog breeds the cropped image is resemble.
cropped image includes only one detected face, so no need to run the whole process in improved_predict_breed function
Parameters:
img_path: a string-valued file path to an image
Return:
None
"""
iV3_model = iV3(self.img_path)
dog_filters = df(self.img_path)
faces, BGR_img = self.Improved_OpenCV_face_detector()
print('Hello, human!')
dog_filters.apply_snapchat_filter()
iV3_model.show_top5_result()
def detect_face_on_cropped_imgs(self, cropped_imgs):
"""
Detect whether if any faces in an image
Parameters:
img_path: a string-valued file path to an image
Return:
True: if face is detected in image stored at img_path
False: if face is NOT detected in image stored at img_path
"""
for i in range(len(cropped_imgs)):
cropped_img_path = u.save_cropped_imgs(i, cropped_imgs)
self.img_path = cropped_img_path
self.predict_breed_for_human_only()
def process_predict(self):
"""
Predict which dog breeds the image is resemble.
if dog and face are detected on the same image
- predict dog breeds on the detected dog
- crop the detected face area on the image and predict dog breeds on the detected face
if more than one detected faces on the same image
- crop the detected face area on the image and predict dog breeds on the individual detected face
if only one face is detected on the same image
- predict dog breeds on the detected face
if no dog and human are detected
- print "No human. No dog."
Parameters:
None
Return:
None
"""
rn50_model = rn50()
iV3_model = iV3(self.img_path)
dog_filters = df(self.img_path)
faces, BGR_img = self.Improved_OpenCV_face_detector()
dogs = rn50_model.dog_detector(self.img_path)
#if dog and human in the same image, model predicts dog breeds will always based on the dog
#so we have to cropped the human image from the dog
if(dogs != 0):
print('Hello, dog!')
u.show_upload_image(self.img_path)
iV3_model.show_top5_result()
if(len(faces) > 0):
cropped_imgs = u.crop_detected_faces(BGR_img, faces)
self.detect_face_on_cropped_imgs(cropped_imgs)
u.delete_cropped_images()
#if more than one people in the same image, model predicts dog breeds will always show one result
#so we have to crop the human image to individuals
else:
if(len(faces) > 1):
cropped_imgs = u.crop_detected_faces(BGR_img, faces)
self.detect_face_on_cropped_imgs(cropped_imgs)
u.delete_cropped_images()
elif(len(faces) == 1):
print('Hello, human!')
dog_filters.apply_snapchat_filter()
iV3_model.show_top5_result()
else:
print('No human. No dog.')
u.show_test_image(self.img_path)
Step 7: Test Your Algorithm¶
In this section, you will take your new algorithm for a spin! What kind of dog does the algorithm think that you look like? If you have a dog, does it predict your dog's breed accurately? If you have a cat, does it mistakenly think that your cat is a dog?
(IMPLEMENTATION) Test Your Algorithm on Sample Images!¶
Test your algorithm at least six images on your computer. Feel free to use any images you like. Use at least two human and two dog images.
Question 6: Is the output better than you expected :) ? Or worse :( ? Provide at least three possible points of improvement for your algorithm.
Answer: There are some issue on the dog and face detector:
Can not identify how many dogs in a image.
- if there are one more dogs at the same image, it will always predict only one result which may be the average result or the dominant dog occupied in the image.
Possible solution:
Dog detector should have the same function as OpenCV detector which can identify how many faces are detected and provide individual face coordinate. Likewise, dog detector should be able to detect how many dogs in the images and provide the coordinate of the dog objects.Can not predict the dog breeds on human if human and dog at the same image.
- The target of the dog detctor is always on the dog
Solution:
Cropped the human faces from the image and take this cropped image as input to the predict model.
The bounding box of detected faces in OpenCV face detector can't automatically flip or rotate the bounding box to the same angle of the detect faces.
- The coordinates to put dog filters are calculated based on the bounding box. If the position of bounding box
is not align to the face well, the filters will not put on the face image appropriately.
Possible solution:
Facial landmark detection may be able to catch the detected face more correctly.- The coordinates to put dog filters are calculated based on the bounding box. If the position of bounding box
## TODO: Execute your algorithm from Step 6 on
## at least 6 images on your computer.
## Feel free to use as many code cells as needed.
predict_breed('test_imgs/dog.jpg')
predict_breed('test_imgs/dog1.jpg')
predict_breed('test_imgs/people.jpg')
show_bounding_box_4_detected_face('test_imgs/people.jpg')
improved_predict_breed('test_imgs/people.jpg')
predict_breed('test_imgs/albert_einstein.jpg')
predict_breed('test_imgs/red_velvet.jpg')
improved_predict_breed('test_imgs/red_velvet.jpg')
predict_breed('test_imgs/celebrity_dog.jpg')
improved_predict_breed('test_imgs/celebrity_dog.jpg')
predict_breed('test_imgs/human_n_dog.jpg')
improved_predict_breed('test_imgs/human_n_dog.jpg')
predict_breed('test_imgs/Dog_n_Cat.jpg')
improved_predict_breed('test_imgs/human_n_dog2.jpg')
improved_predict_breed('test_imgs/human_n_dog3.jpg')
show_bounding_box_4_detected_face('test_imgs/human_n_dog3.jpg')
predict_breed('test_imgs/Golden_Cocker_Retriever.jpg')
predict_breed('test_imgs/dogs.jpg')
predict_breed('test_imgs/dogs2.jpg')
predict_breed('test_imgs/cat.jpg')
predict_breed('test_imgs/people_cartoon.png')
predict_breed('test_imgs/dog_cartoon.jpg')
predict_breed('test_imgs/cocker_spaniel_clipart_sketch.jpg')
improved_predict_breed(train_files[0])
Web App¶
Please check dog breed web_app folder.
from IPython.display import Image
Image(filename='web_app_snapshot.png')
Please download your notebook to submit¶
In order to submit, please do the following:
- Download an HTML version of the notebook to your computer using 'File: Download as...'
- Click on the orange Jupyter circle on the top left of the workspace.
- Navigate into the dog-project folder to ensure that you are using the provided dog_images, lfw, and bottleneck_features folders; this means that those folders will not appear in the dog-project folder. If they do appear because you downloaded them, delete them.
- While in the dog-project folder, upload the HTML version of this notebook you just downloaded. The upload button is on the top right.
- Navigate back to the home folder by clicking on the two dots next to the folder icon, and then open up a terminal under the 'new' tab on the top right
- Zip the dog-project folder with the following command in the terminal:
zip -r dog-project.zip dog-project - Download the zip file by clicking on the square next to it and selecting 'download'. This will be the zip file you turn in on the next node after this workspace!